vdr-frontend 100% cpu

  • Hi!


    I have been having this weird random problem. VDR becomes very unresponsive and starts skipping frames etc, then when I use "top" I see that python process is using 100% of CPU..


    This seems to be the vdr-frontend. There is a python script in /etc/init/vdr-frontend.conf


    If I kill the python process, vdr-frontend dies but becomes back soon. The high cpu load is gone, but will come back later.


    This is yavdr 0.5.


    I'm tempted to remove the whole frontend, why is it needed? I'm not using xbmc and my VDR box is running 24h, it would probably work just fine with VDR just started directly.. Is vdr-frontend doing something useful?

  • I guess you are looking for a recording and streaming server, right ?
    Open the webinterface and set the frontend to headless - mode.


    If not then: Do you have a VDPAU capable nvidia grafic.card installed ?
    Otherwise yavdr is not suitable for you...


    Have you updated after the installation ?


    Code
    sudo apt-get update
    sudo apt-get dist-upgrade



    KR,
    Joe

  • Please post the /var/log/upstart/vdr-frontend.log so I can have a look what is going wrong.

    yaVDR-Dokumentation (Ceterum censeo enchiridia esse lectitanda.)

  • Hi,


    most of the log seems to be quite uninteresting, but when I kill python I get this:


    /proc/self/fd/9: line 399: 3262 Terminated python - <<END


    line 399 is in eventlircd event handling.


    I recently posted about my eventlircd problems, but no one responded.. I guess this is related! eventlircd starts loosing remote keypresses when VDR is started or restarted. "restart eventlircd" fixes it.


    However, I now notice that after eventlircd is restarted, vdr-frontend cpu goes to 100%!


    So now I have a choice.. :)
    1) If eventlircd looses half of my keypresses, vdr-frontend works ok.
    2) If eventlircd works fine, vdr-frontend starts using 100% cpu.


    Any ideas?

  • I uploaded a newer version (in yavdr-utils) - could you do an dist-upgrade and test it again?

    yaVDR-Dokumentation (Ceterum censeo enchiridia esse lectitanda.)

  • Hi, thanks for your quick support!


    Now, I can do "restart eventlircd" and vdr-frontend doesn't start 100% cpu. So it's "fixed", i guess!


    However, now the vdr remote always looses half of the keypresses. :) restart eventlircd doesn't fix it.


    I'm thinking about trying the remote plugin, it probably would work just fine for me because I don't need to use the remote in xbmc? Just vdr..

  • Which type of remote an receiver do you use?

    yaVDR-Dokumentation (Ceterum censeo enchiridia esse lectitanda.)

  • I have TBS6922 and TBS6891 dvb-s2 cards, with built-in IR receivers.. They appear in /dev/input/ directory.


    This worked fine with my previous machine, directly with eventlircd without any configuring. But I changed the motherboard, and then the keypresses started disappearing. The software should be the same, I simply changed the old harddrive to the new machine.

  • Is there any difference in the output of irw if the vdr-frontend is stopped compared to a running frontend script?
    If you stop eventlircd and use evtest to listen to the input device directly - does it show all the key presses correctly?

    yaVDR-Dokumentation (Ceterum censeo enchiridia esse lectitanda.)

  • evtest always shows all the keypresses, if eventlircd is stopped.


    irw shows the same that's received by VDR; if the keypress is missed, it won't appear on irw either..


    Having vdr-fronted stopped or started makes no difference! Keys are missed..


    In the meantime, I now tried remote plugin and it seems to work, vdr works nicely and keys are not missed.. However vdr-frontend now floods the syslog complaining that it can't find eventlircd socket. :) Perhaps it should check the socket at start, and remove eventlircd monitoring if eventlircd is not running?

  • evtest always shows all the keypresses, if eventlircd is stopped.


    Are both ir-receivers connected to your DVB-cards? This would irritate eventlircd because the expected key status signals get mixed up.

    yaVDR-Dokumentation (Ceterum censeo enchiridia esse lectitanda.)


  • Are both ir-receivers connected to your DVB-cards? This would irritate eventlircd because the expected key status signals get mixed up.

    No, only single one connected.


    I can connect the ir-receiver to any DVB-card, and the behaviour is the same.. Weird.


    Anyway.. I have this now pretty much working with remote plugin.. Of course it would be better to have it working as expected, but I can live with this solution. :)

  • I can connect the ir-receiver to any DVB-card, and the behaviour is the same.. Weird.


    Eventlircd expects a certain order of key states when forwarding keypresses - could you paste the output of evtest for some keypresses?

    yaVDR-Dokumentation (Ceterum censeo enchiridia esse lectitanda.)

  • Hi,


    sorry for the small delay, had other things to do. :)


    I'm really clueless what's happening! I stopped eventlircd, and run evtest. For some reason, VDR still receives keypresses! Where do they come from? Perhaps that's the problem, keypresses coming from both eventlircd and some other route?


    Here is the evtest result..


  • For some reason, VDR still receives keypresses! Where do they come from?


    From the kernel input device (but this time not over the lirc-socket eventlircd provides for the vdr but via the X-server).

    Perhaps that's the problem, keypresses coming from both eventlircd and some other route?


    No, if eventlircd binds a device it sets a lock preventing other processes from reading the device.


    The output of evtest does look quite normal, too - always a value 1 (=key down) followed by a value0 (=key up).

    Code
    Event: time 1363279804.324859, type 1 (EV_KEY), code 108 (KEY_DOWN), value 1
    Event: time 1363279804.324860, -------------- SYN_REPORT ------------
    [...]
    Event: time 1363279804.433170, -------------- SYN_REPORT ------------
    Event: time 1363279804.684817, type 1 (EV_KEY), code 108 (KEY_DOWN), value 0


    I have no idea why eventlircd should throw away key presses (unless someone reenabled it's repeat-filter in /etc/init/eventlircd.conf)

    yaVDR-Dokumentation (Ceterum censeo enchiridia esse lectitanda.)

  • Hmm! Something interesting has happened..


    As I said, I have two TBS6922 and one TBS6891 card. I have tried IR receiver on all three, with similar results.


    But now I try it, and TBS6922's have the bug and TBS6981 doesn't! It works nicely with eventlircd now, but only with TBS6981.. But that's enough, of course.


    But on my previous machine it worked fine with TBS6922 as well. And yesterday it didn't work with TBS6981, but now does. :)


    I better not touch anything now. :)

Jetzt mitmachen!

Sie haben noch kein Benutzerkonto auf unserer Seite? Registrieren Sie sich kostenlos und nehmen Sie an unserer Community teil!